def test_init_game(self):
game = Game()
state, player_id = game.init_game()
test_id = game.get_player_id()
self.assertEqual(test_id, player_id)
self.assertIn('call', state['legal_actions'])
self.assertIn('raise', state['legal_actions'])
self.assertIn('fold', state['legal_actions'])
def test_payoffs(self):
game = Game()
np.random.seed(0)
for _ in range(5):
game.init_game()
while not game.is_over():
legal_actions = game.get_legal_actions()
action = np.random.choice(legal_actions)
game.step(action)
payoffs = game.get_payoffs()
total = 0
for payoff in payoffs:
total += payoff
self.assertEqual(total, 0)
def __init__(self, config):
''' Initialize the Limitholdem environment
'''
self.game = Game()
super().__init__(config)
self.actions = ['call', 'raise', 'fold', 'check']
self.state_shape = [72]
with open(
os.path.join(rlcard3.__path__[0],
'games/limitholdem/card2index.json'),
'r') as file:
self.card2index = json.load(file)
class LimitholdemEnv(Env):
''' Limitholdem Environment
'''
def __init__(self, config):
''' Initialize the Limitholdem environment
'''
self.game = Game()
super().__init__(config)
self.actions = ['call', 'raise', 'fold', 'check']
self.state_shape = [72]
with open(
os.path.join(rlcard3.__path__[0],
'games/limitholdem/card2index.json'),
'r') as file:
self.card2index = json.load(file)
def _get_legal_actions(self):
''' Get all leagal actions
Returns:
encoded_action_list (list): return encoded legal action list (from str to int)
'''
return self.game.get_legal_actions()
def _extract_state(self, state):
''' Extract the state representation from state dictionary for agent
Note: Currently the use the hand cards and the public cards. TODO: encode the states
Args:
state (dict): Original state from the game
Returns:
observation (list): combine the player's score and dealer's observable score for observation
'''
extracted_state = {}
legal_actions = [self.actions.index(a) for a in state['legal_actions']]
extracted_state['legal_actions'] = legal_actions
public_cards = state['public_cards']
hand = state['hand']
raise_nums = state['raise_nums']
cards = public_cards + hand
idx = [self.card2index[card] for card in cards]
obs = np.zeros(72)
obs[idx] = 1
for i, num in enumerate(raise_nums):
obs[52 + i * 5 + num] = 1
extracted_state['obs'] = obs
if self.allow_raw_data:
extracted_state['raw_obs'] = state
extracted_state['raw_legal_actions'] = [
a for a in state['legal_actions']
]
if self.record_action:
extracted_state['action_record'] = self.action_recorder
return extracted_state
def get_payoffs(self):
''' Get the payoff of a game
Returns:
payoffs (list): list of payoffs
'''
return self.game.get_payoffs()
def _decode_action(self, action_id):
''' Decode the action for applying to the game
Args:
action id (int): action id
Returns:
action (str): action for the game
'''
legal_actions = self.game.get_legal_actions()
if self.actions[action_id] not in legal_actions:
if 'check' in legal_actions:
return 'check'
else:
return 'fold'
return self.actions[action_id]
def _load_model(self):
''' Load pretrained/rule model
Returns:
model (Model): A Model object
'''
return models.load('limit-holdem-rule-v1')
def get_perfect_information(self):
''' Get the perfect information of the current state
Returns:
(dict): A dictionary of all the perfect information of the current state
'''
state = {}
state['chips'] = [
self.game.players[i].in_chips for i in range(self.player_num)
]
state['public_card'] = [c.get_index() for c in self.game.public_cards
] if self.game.public_cards else None
state['hand_cards'] = [[
c.get_index() for c in self.game.players[i].hand
] for i in range(self.player_num)]
state['current_player'] = self.game.game_pointer
state['legal_actions'] = self.game.get_legal_actions()
return state
def test_get_player_num(self):
game = Game()
player_num = game.get_player_num()
self.assertEqual(player_num, 2)
def test_step_back(self):
game = Game(allow_step_back=True)
game.init_game()
self.assertEqual(game.step_back(), False)
index = 0
previous = None
while not game.is_over():
index += 1
legal_actions = game.get_legal_actions()
if index == 2:
result = game.step_back()
now = game.get_player_id()
if result:
self.assertEqual(previous, now)
else:
self.assertEqual(len(game.history), 0)
break
previous = game.get_player_id()
action = np.random.choice(legal_actions)
game.step(action)
def test_step(self):
game = Game()
# test raise
game.init_game()
init_raised = game.round.have_raised
game.step('raise')
step_raised = game.round.have_raised
self.assertEqual(init_raised+1, step_raised)
# test call
game.init_game()
init_not_raise_num = game.round.not_raise_num
game.step('call')
step_not_raise_num = game.round.not_raise_num
self.assertEqual(init_not_raise_num+1, step_not_raise_num)
# test fold
game.init_game()
game.step('fold')
self.assertTrue(game.round.player_folded)
# test check
game.init_game()
game.step('call')
game.step('check')
self.assertEqual(game.round_counter, 1)
# test play 4 rounds
game.init_game()
for i in range(19):
if (i+1) % 5 == 0:
game.step('call')
else:
game.step('raise')
self.assertEqual(game.is_over(), False)
game.step('call')
self.assertEqual(game.is_over(), True)
# Test illegal actions
game.init_game()
with self.assertRaises(Exception):
game.step('check')
# Test the upper limit of raise
game.init_game()
for _ in range(4):
game.step('raise')
legal_actions = game.get_legal_actions()
self.assertNotIn('raise', legal_actions)
def test_get_action_num(self):
game = Game()
action_num = game.get_action_num()
self.assertEqual(action_num, 4)